In the Central Vein Occlusion Study (CVOS), the majority of eyes with central retinal vein occlusion (CRVO) and extensive intraretinal hemorrhages (IRH) at baseline were later classified as ischemic.1 A recent study showed an association between the pattern of IRH in the periphery and retinal capillary nonperfusion in eyes with CRVO.2 It is plausible that eyes with CRVO presenting with a fovea-involving IRH may be at an increased risk for macular ischemia and associated vision loss.
Vascular endothelial growth factor (VEGF) is an important contributor to the development of macular edema in CRVO,3, 4 and treatment with anti-VEGF agents have become the standard of care in these eyes. Intraretinal hemorrhage clears more rapidly following suppression of VEGF; however, the exact mechanism by which anti-VEGF agents facilitate IRH resolution has not been well described.5
Visual and anatomical outcomes in CRVO eyes with fovea-involving IRH treated with anti-VEGF therapy, however, have not been previously reported. These eyes have been excluded from several lead clinical studies of agents targeting VEGF.4–10 These exclusions make it difficult to extrapolate results from these studies to CRVO eyes with fovea-involving IRH. The purpose of our study is to better understand the impact and implication of fovea-involving IRH in eyes with acute, treatment-naïve CRVO.
Patients and Methods
A retrospective review of medical records was performed on all patients diagnosed with acute, treatment-naïve CRVO at the Duke Eye Center from January 1, 2009, to June 30, 2016. This study was conducted with the approval of the Duke University Medical Center Institutional Review Board and in accordance with the principles of the Declaration of Helsinki. Data collection was compliant with the Health Insurance Portability and Accountability Act (HIPAA). Only patients who were treatment-naïve with CRVO onset 3 months or less prior to presentation and who had 12 months or more of follow-up were included. Patients without baseline spectral-domain optical coherence tomography (SD-OCT) and color fundus photographs were excluded.
Evaluation for Presence of a Fovea-Involving Intraretinal Hemorrhage
Grading From Color Fundus Photographs (Clinical Method): Baseline fundus photographs were graded for the presence or absence of a fovea-involving IRH by three independent graders (SPY, FB, MGJ). A patient was considered to have fovea-involving IRH if any portion of an IRH involved the fovea. The fovea was approximated to be the area, roughly a disc diameter in size, centered in the area of deepest xanthophyll pigmentation, when visible, approximately 2.5 disc radii temporal to the center of the optic disc on fundus photographs.11 Discordant grades were adjudicated by a fourth grader (AST). Those with a fovea-involving IRH were classified as “group 1” and those without such a hemorrhage were classified as “group 2.”
Grading From Multimodal Imaging: Multimodal imaging was used to validate the use of color fundus photograph grading as a method of accurately detecting foveal IRH. To this end, all images were further analyzed by a single grader (AST). The fundus photographs and near infrared (NIR) SD-OCT images acquired using automatic retinal tracking, with a horizontal and vertical line scan passing through the foveal center, were overlaid. The point of intersection of the vertical and horizontal line scans was determined to be the foveal center. A circle, 1.5 mm in diameter centered on the foveal center and therein representing the fovea, was then drawn (Figure 1). The presence of IRH within any part of this circle was graded as positive for fovea-involving IRH. Next, agreement between presence or absence of fovea-involving IRH using multimodal imaging versus the clinical method was analyzed.
Grading for presence of fovea-involving intraretinal hemorrhage (IRH) using multimodal imaging. The color or pseudocolor fundus photo (A) was overlaid with near-infrared images with a horizontal (B) and vertical (C) line scan centered on the fovea. The intersection of the line scans was identified as the foveal center (D). A circle 1.5 mm in diameter centered on this foveal center was drawn that represented the outer boundary of the fovea. Presence of any portion of an IRH within this circle was graded as fovea-involving.
Data on study patients were collected retrospectively, capturing the initial presenting visit and final follow-up visit. Individual patient charts were reviewed using the EPIC medical record system, and data were entered into a password-protected Microsoft Excel (version 14; Microsoft, Redmond, WA) spreadsheet. Demographic information and medical history were documented at baseline. Clinical examination findings were recorded from the baseline and final follow-up visits. SD-OCT scans were graded for the presence of CME and central subfield thickness (CST). Treatment data collected included the number and type of anti-VEGF injections and application of panretinal laser photocoagulation (PRP). Corrected VA was measured using the Early Treatment Diabetic Retinopathy Study (ETDRS) protocol and then converted to logarithm of the minimum angle of resolution (logMAR) scale.
All statistical analyses were performed using Stata software Version 10.0 (Stata Corp 2007 Stata Statistical Software: Release 10; Stata Corp, College Station, TX). Comparisons between groups were made using the independent samples t-test for parametric variables and the Mann-Whitney U test for nonparametric variables. For categorical variables, comparison between groups was made using the Chi-square test for a larger sample size (n > 5) and Fisher's exact test for a smaller sample size (n < 5). Simple and multiple linear regression models were used to examine the association between final VA and the previously mentioned variables. The estimated beta coefficients and their 95% confidence intervals from the regression models were presented. All P values were nominal, and a value of less than .05 was considered statistically significant. Values are presented as mean ± standard error of mean (SEM) unless otherwise specified.
Patient Demographics and Baseline Characteristics
At presentation, 115 (39.8%) of the 289 patients with CRVO had fovea-involving IRH. Baseline demographic characteristics for patients with (group 1) or without (group 2) fovea-involving IRH are listed in Table 1. The mean age was 66.3 years ± 1.4 years and 63.3 years ± 1.3 years for groups 1 and 2, respectively. Approximately 50% of patients in group 1 and 60% in group 2 were female. Differences existed in the distribution of race with a significantly higher percentage of African-Americans in group 1. Patients in group 1 were also more likely to have a positive smoking history.
Demographics and Clinical Characteristics of Patients Presenting With or Without Fovea-Involving Intraretinal Hemorrhage at the Baseline Visit
Use of antiplatelet agents was more common for patients in group 1 (52% vs. 37%; P = .004). Forty-seven percent (n = 49) receiving low-dose aspirin (81 mg) and 57.1% (n = 8) receiving high-dose aspirin (325 mg) had a fovea-involving IRH. Hemorrhage rates were lower for clopidogrel (40%, n = 6) in comparison to both high- and low-dose aspirin. Both groups reported similar use of anticoagulants; six of 10 (60%) patients on warfarin, one of three (33.3%) on apixaban, one of two (50%) on dabigatran, and none of the three patients on rivaroxaban therapy developed foveal IRH. As the number of patients on these agents was small, the association of these antiplatelet and anticoagulant agents with foveal hemorrhage could not be reliably determined.
Mean baseline logMAR (Snellen equivalent) visual acuity (VA) was worse for patients in group 1 (1.16 ± 0.06 [20/289] vs. 0.86 ± 0.06 [20/145]; P = .001) (Table 1). A higher percentage of eyes in group 1 presented with CME at baseline (85.2% vs. 62.6%; P < .001), and these eyes also had a significantly higher CST in comparison to eyes in group 2 (610.38 μm ± 35.87 μm vs. 434.99 μm ± 21.64 μm; P < .001).
Comparison of Visual and Anatomical Outcomes for Eyes With and Without Fovea-Involving IRH
The mean follow-up duration was similar for both groups (46.1 vs. 34.9 months; P = .38). Ten (8.7%) patients in group 1 and 22 (12.6%) in group 2 (P = .31) developed new onset RVO (CRVO or branch retinal vein occlusion) in the fellow eye by the final follow-up visit.
The final logMAR (Snellen equivalent) VA was comparable for subjects in groups 1 and 2 (1.24 logMAR ± 0.09 logMAR [20/348] vs.1.02 logMAR ± 0.08 logMAR [20/209]; P = .08) (Table 2). Changes from baseline in mean VA at the final follow-up visit were also comparable (0.11 logMAR ± 0.05 logMAR vs. 0.08 logMAR ± 0.08 logMAR). However, a larger percentage of eyes in group 2 had final VA better or equal to 20/40 compared to those in group 1 (17% vs. 24%; P < .001). A significantly higher percentage of eyes in group 1 had final improvement in vision of 2 or more lines (35.7% vs. 17.2%; P = .002) and 3 or more lines (31.3% vs. 10.3%; P < .001).
Clinical Characteristics of Central Retinal Vein Occlusion Patients With or Without Fovea-Involving Intraretinal Hemorrhage at the Final Follow-up Visit
The percentage of patients with CME at the final follow-up visit was similar for the two groups (33.9% vs. 29.9%, P = .76). The mean CST at the final follow-up visit was also comparable (291.90 μm ± 18.35 μm vs. 327.59 μm ± 18.36 μm; P = .17) (Table 2). There was a significantly greater decrease in mean CST at the final follow-up visit for subjects in group 1 in comparison to group 2 (325.5 μm ± 42.6 μm vs.136.9 μm ± 27.8 μm; P < .001).
Baseline Predictors of Final Visual Outcome
Age, CRVO duration at presentation, baseline VA, baseline CST, and presence of foveal IRH were analyzed as factors predictive of the final visual outcome. Univariate linear regression analysis found age (β = −0.40; P = .006), baseline VA (β = 0.72; P < .001), baseline CST (β = −0.02; P = .009), and foveal IRH (β = −8.05; P = .05) as factors predictive of the final visual outcome (Table 3.) Multiple linear regression analysis detected baseline VA (β = 0.75; P < .001) as the only factor predictive of the final visual outcome (Table 3).
Baseline Predictors of Final Visual Outcome Using Univariate and Multiple Linear Regression Models
Baseline Predictors of Treatment Burden With Anti-VEGF Agents
Patients in group 1, on average, received a significantly greater number of intravitreal anti-VEGF injections during the first year of follow-up (7.80 ± 0.40 vs 5.20 ± 0.40; P = .001). A significantly higher percentage of patients in group 1 received PRP during follow-up (32.2% vs. 13.8%; P < .001). Univariate and multiple linear regression models were used to determine baseline predictors of treatment burden with anti-VEGF agents as determined by the number of anti-VEGF injections received during the first year of follow-up (Table 4). Disease duration (β = −0.15; P = .03), VA (β = −0.24; P = .001), CST (β = 0.26; P < .001), and foveal IRH (β = 0.17; P = .02) were baseline factors predictive of treatment burden.
Baseline Predictors of Treatment Burden With Anti-Vascular Endothelial Growth Factor Using Univariate and Multiple Linear Regression Models
The inter-reader Pearson correlation coefficient for evaluation of a fovea-involving IRH was 0.97. The Pearson correlation coefficient for the presence or absence of a fovea-involving IRH using the clinical method and the multimodal imaging method was excellent (R = 1.00).
Intraretinal hemorrhage is a well-described ophthalmoscopic feature in eyes with acute CRVO; however, at present, our understanding of how fovea-involving IRH specifically affects the visual and anatomical outcomes and treatment burden in eyes with CRVO is limited. In the CVOS trial, the majority of CRVO eyes with extensive, diffuse IRH at baseline was later classified as ischemic.1 Intraretinal hemorrhage obscuring the fovea was an exclusion criteria for the CRUISE trial4 and for follow-up studies HORIZON8 and RETAIN.10 The SHORE trial also excluded eyes with visually significant foveal IRH.9 Although the COPERNICUS and GALILEO trials did not exclude eyes with fovea-involving IRH, outcomes for this subset of eyes were not separately reported in these studies.6,7 The purpose of our study is to better understand the impact and implication of fovea-involving IRH in eyes with acute, treatment-naïve CRVO. To the best of our knowledge, this is the first study that has specifically reported outcomes in CRVO eyes with foveal-involving IRH.
Demographic characteristics of the two groups were similar with the exception of race, with an increased proportion of African Americans having fovea-involving IRH. This may be partially explained by the higher use of antiplatelet medications among African Americans in comparison to Caucasians in our study cohort. Our study showed that patients in group 1 were more likely to be on antiplatelet therapy compared to those in group 2, suggesting antiplatelet medication use as a potential risk factor for the development of fovea-involving IRH. In a large prospective study of 686 patients with CRVO or hemi-CRVO, Hayreh et al. described more marked intraretinal hemorrhage among aspirin users compared to nonusers.12 Although Hayreh's study did not differentiate between low- and high-dose aspirin, results from our study reveal that a greater proportion of patients taking high-dose aspirin developed fovea-involving IRH in comparison to low-dose aspirin, suggesting a dose-related increase in risk.
Comparison of visual and anatomical outcomes in our study showed that although patients in group 1 presented with significantly worse VA at baseline, the final visual outcome between the two groups was comparable. The poorer baseline VA in group 1 may be explained by the presence of fovea-involving IRH and more marked macular edema. Although macular ischemia was difficult to assess at baseline for group 1 patients due to the presence of IRH, we know from the CVOS that patients with extensive IRH throughout the fundus are more likely to be classified as ischemic over time.1 The changes from baseline in mean VA at the final follow-up visit were similar in both groups, although a significantly higher percentage of eyes in group 1 had 2 or more and 3 or more lines' improvement in VA. This may be explained by the worse baseline VA in eyes with fovea-involving IRH, therefore providing a greater opportunity for visual improvement and/or due to a ceiling effect in group 2 given the better baseline VA. Although our study is the first to report visual outcomes in CRVO eyes with fovea-involving IRH, the previously described study by Hayreh et al. corroborates our study results.12 Furthermore, our study demonstrated that baseline visual acuity was the strongest predictor of final VA in patients with fovea-involving IRH. This is similar to what has been previously described by studies reporting visual outcomes in CRVO eyes without fovea-involving IRH treated with anti-VEGF agents.13,14
Eyes in group 1 received a significantly greater number of anti-VEGF injections during the first year of follow-up. These eyes likely had higher levels of VEGF due to more widespread retinal ischemia, which is suggested by a greater baseline CST. The greater baseline CST in eyes with foveal intraretinal hemorrhage could likely explain the increased number of anti-VEGF injections received in this group. Interestingly, results from the multiple linear regression model showed foveal IRH to be an independent predictor of treatment burden after controlling for CST. Although subjects with fovea-involving IRH presented with a significantly worse CST, the final anatomical outcome after treatment with anti-VEGF agents was similar to those without such hemorrhage. Eyes in group 1, however, did have a greater treatment burden, which can have potential cost implications and necessitate more frequent follow-up. Larger prospective studies are needed to establish whether these differences are consequential in patient care.
The main limitations of this study include its retrospective study design, making it difficult to control for the type and frequency of anti-VEGF treatments received during follow-up. Although we stratified outcomes based on the presence or absence of a fovea-involving IRH, we did not quantify the severity of IRH throughout the fundus in our study population. We also do not report data on retinal ischemia. In our study, macular nonperfusion and foveal avascular zone size were difficult to assess on fluorescein angiography due to the presence of IRH. However, patients in group 1 had higher rates of neovascular sequelae, suggesting greater underlying retinal ischemia. Although it is conceivable that fovea-involving IRH could be a clinical sign of underlying macular ischemia, future studies correlating presence or absence of foveal IRH with retinal vessel density using OCT angiography may help clarify this further. The sample size, although large overall, limited the ability to perform a strong subgroup analysis comparing outcomes between the different types of anti-VEGF agents and treatment frequency. Referral bias may also be present, as all patients were seen at a single tertiary care medical center, and our patient population may differ from community-based practices.
Our study compares visual and anatomical outcomes and treatment burden in treatment-naïve eyes with acute CRVO with and without fovea-involving IRH. This study concludes that although CRVO eyes with accompanying fovea-involving IRH present with a worse VA and more marked macular edema at baseline, the final visual and anatomical outcomes are comparable to eyes without such hemorrhage. Those with fovea-involving IRH do, however, require a greater number of anti-VEGF injections to obtain similar visual and anatomical outcomes to those without foveal IRH.